LF-Ant: A Bio-inspired Cooperative Cross-layer

Design for Wireless Sensor Networks

Marcelo Portela Sousa, Waslon Terllizzie Araujo Lopes and Marcelo S. AlencarInstitute for Advanced Studies in Communications (Iecom), Campina Grande, PB, Brazil,

Federal University of Campina Grande (UFCG), Campina Grande, PB, Brazil.

E-mails: {marcelo.portela,waslon,malencar}@ieee.org

Abstract—In this paper, the authors propose the LF-Antprotocol for multi-hop wireless sensor networks, a cross-layerdesign inspired in the organized and collaborative behaviour ofnatural ants. At the network layer, the heuristic information ismodelled by a fuzzy inference system to assist a cluster-headelection and the routing process. A resultant relaying thresholdis combined with an adaptive invoking of cooperative modulationdiversity, at link and physical layers. Simulation results show theperformance enhancement in network lifetime and packet lossrate, compared with another cross-layer cooperative system.

I. INTRODUCTION

Wireless sensor networks (WSN) increased the research

interest in developing efficient techniques for the monitoring

of specific regions. Moreover, the constraints in terms of power

supply require protocols with efficient use of energy resources

and autonomous infrastructure deployment.

Cluster-based protocols are successful methods for energy

saving. In each group, a cluster-head is elected as coordinator

of that cluster, which collects data from other nodes, aggre-

gates and reports it to the sink node. In clustered schemes,

the cluster-head election process is a fundamental issue and

impacts significantly in the network energy consumption.

The clustering inspired by the organized behaviour of social

ants, with the use of Ant Colony Optimization (ACO) have

been developed to enhance the performance of WSNs [1].

However, that proposed schemes are based on single-hop

networks, which waste energy and limit the transmission

distance range [2].

In ACO systems, the information collected by the ants in

the search process is stored in pheromone trails, τ . The arcs

also have a priori information, η, which is heuristic about

the problem definition. If η represents a cost function related

to distance, the imprecision in measurement can degrade the

performance of the system [3].

Apart from the limited resources, the fading caused by mul-

tipath in wireless channels affect the quality of transmission

and increases the Packet Loss Rate (PLR). The use of truncated

Automatic Repeat Request (ARQ) decreases the PLR, but

depending on the relation between the maximum number of re-

transmissions allowed and the channel quality, it may degrade

the network lifetime due the required retransmissions [4].

Cooperative modulation diversity (CMD) can solve that

tradeoff, by combating the fading effects without incurring

waste of bandwidth or energy. Different of other recently

proposed cooperative schemes, as SCA with LEACH, which

spends too much energy in introducing redundancy by a

space-time block coding, CMD rotates the angle of the sig-

nal constellation, and interleaves the transmitted component

symbols [5]–[7].

This paper presents the LF-Ant (Linguistic Fuzzy Ant)

protocol, a cooperative and cross-layer design for multi-hop

wireless sensor networks, inspired by the behaviour of ants.

LF-Ant intends to increase the network lifetime and decrease

the packet loss rate of WSNs. The goals are reached by an

optimal election of cluster-heads, at network layer, and by the

control of possible further CDM retransmissions, at link and

physical layers. A relaying threshold and a truncated ARQ

process guide the cross-layer operation. Simulations compare

the LF-Ant performance with another cross-layer cooperative

design and attest its performance enhancement.

The major contribution of the paper is the proposing of

a novel clustering protocol, which also uses another novel

proposed concepts. The fuzzy heuristic information deals with

measurements uncertainties of wireless channels, by a fuzzy

inference system, and enhances the traditional ACO usage,

which is based on crisp logic. Furthermore, the vice cluster-

head entity is proposed to support the cooperation of nodes,

with the cooperative modulation diversity, and to reduce the

number of required retransmissions in the ARQ system.

The remaining of the paper is organized as follows: Sec-

tion II provides an overview of a classical ACO system oper-

ation related to a generic routing problem in communication

networks, the AntNet system. Section III models the proposed

clustering protocol, with the use of fuzzy heuristic information.

The operation of cooperative modulation diversity is described

in Section IV. Simulation results are discussed in Section V

and the conclusions are summarized in Section VI.

II. A BRIEF OVERVIEW OF THE ANTNET OPERATION

In AntNet, each ant searches for a minimum cost path

between a pair of nodes, i and d [8]. If an ant κ is in node i, it

hops to j, in accordance with a decision rule that is a function

of the ant’s memory, Mκ, and of the local ant-routing table,

Ai. That table is obtained by a composition of the pheromone

trails, τijd, and of the heuristic information, ηijd. Once the ant

κ has completed a path, it deposits an amount of pheromone,

∆τκ, proportional to the goodness of the path it built. In this

way, after reaching its destination node, the ant moves back

2011 IEEE 22nd International Symposium on Personal, Indoor and Mobile Radio Communications

978-1-4577-1348-4/11/$26.00 ©2011 IEEE 289

to its source node, along the same path, and increases the

pheromone intensity, modelled by the Assignment 1:

τijd ← τijd +∆τκ. (1)

To prevent a premature convergence to non-optimal so-

lutions, the pheromone of outgoing connections evaporates,

indicated by the Assignment 2:

τijd ←τijd

(1 + ∆τκ), ∀j ∈ Ni, (2)

in which Ni is the set of node i neighbours. The relation

between Ai, and the ant’s decision rule, pκijd, is given by:

pκijd =

ωτijd + (1 − ω)ηijdω + (1− ω)(|Ni| − 1)

, if j /∈Mκ,

0, if j ∈Mκ,(3)

in which ω ∈ [0, 1] is a weighting factor between τijd and

ηijd, and the denominator is a normalization term. The ant’s

memory, Mκ, indicates the set of nodes visited by the ant,

and its use can avoid the occurrence of loops.

III. LF-ANT: A NOVEL CLUSTERING PROTOCOL

The proposed clustering protocol, LF-Ant, is based on the

behaviour of ants that need to find optimal paths from the

source to a final destination. The main goal is the opti-

mal election of cluster-heads, in each round, and control of

possible further cooperative retransmissions. LF-Ant modifies

the classical modelling of the ACO system, AntNet, since it

translates the representation of vertices into edges, as well as

the representation of edges into vertices. That is, each sensor

node, s (vertex), in the network that uses the LF-Ant protocol

can be seen as a path (edge), ijd, in the AntNet system. Then,

the election of the best sensor node as a cluster-head, by the

cluster, is equivalent to a choice of the best path from a source

to a final destination, by an ant.

The operation of LF-Ant starts with the random deployment

of artificial ants in the monitored region. Each cluster receives

an ant, κ, that indicates the respective first elected cluster-head.

In the next elections, each cluster node runs decision rules and

generates special values, denoted chance.

For each sensor node, the respective ant, κ, travels to the

final destination and moves back to the nest increasing the

pheromone intensity. In the sensor network domain, this is

equivalent to each sensor node running Assignment 1. The

update variable, ∆τκ, indicates the quality of the chosen path

by the ant. This can measure how good was the operation

performance of the sensor node in the previous round. In LF-

Ant, the update variable is:

∆τκ =εs · ρsζs · Γs

, (4)

in which εs is the residual energy of the node, ζs is the

energy consumed by the node in the previous round, Γs is

the total number of transmissions realized by the node in

the previous round and ρs = 2, if the previous transmitted

packet was successfully recovered by the next destination

node. Otherwise, if even with retransmissions the previous

packet was not correctly recovered, ρs = 1. In (3), ω = 0.5,

because it guarantees the equal weighing between τijd and

ηijd. Those values were determined empirically, i.e., they

optimize the simulation results.

After the trail update, all the sensor nodes run Assignment 2,

which corresponds to the pheromone trails evaporation. The

next step is the processing of the heuristic information, given

by the output of a fuzzy inference system, explained in the next

section. By combining of τs and ηs, the sensor node calculates

the value of the decision rule, according to Formula 3. This

value indicates the probability of a path attract ants, and

equivalently, indicates the probability of a sensor node elect

itself as a cluster-head.

The nodes state the decision rule value as the variable

chance. Each node advertises a message for the other can-

didates, with that value attached, and waits messages from

other nodes. If its chance is higher than the chance from

other nodes, the sensor node advertises a cluster-head

message, which means that the sensor node elected itself a

cluster-head. If a node which is not a cluster-head receives

a cluster-head message, it selects the closest cluster-

head as its coordinator and sends a message to join that cluster.

The sensor node which has the next higher value of chance

(the second place in the election) becomes a relay candidate,

which, in a possible retransmission stage, may become a vice

cluster-head and cooperate in a diversity scheme. That node

reaches a relaying threshold, and the only chance greater in

that cluster is the chance of the elected cluster-head (the first

place in the election).

A. Fuzzy Heuristic Information

In ACO algorithms, the heuristic information represents

a local information which does not depend on the quality

of previous iterations. In Ant System, an algorithm that

optimizes solutions to the travelling salesman problem (TSP),

ηij = 1/Jij , in which Jij represents the distance between

cities i and j [3]. In AntNet, the heuristic information is given

by [8]

ηij = 1−qij

∑

l∈Niqil

, (5)

in which qil is the queue length (in bits to be sent) of the

link that connects the node i to its neighbour j. The set of

neighbours of node i is given by Ni. In LF-Ant, the heuristic

information ηs, represented by the variable eta, relates two

other variables: local_distance and CH_dispersion.

local_distance is the sum of distances between the

candidate node and other nodes within a specific radius of

transmission. The greater the sum, the higher the energy to

transmit the sensed data to the candidate node. The operation

of this variable was proposed in the CHEF protocol [9].

However, if there are few nodes within a specific radius of

transmission, the sum of distances between the candidate node

and other nodes can be small. In this case one may infer,

erroneously, that the node energy consumption is lower than

in the case in which the nodes, in a higher number, are located

closer to the candidate node. The proposed scheme overcomes

290

TABLE IFUZZY IF-THEN RULES USED IN LF-ANT PROTOCOL.

RuleIF THEN

local_distance CH_dispersion eta

1 Close Far Very High

2 Medium Far High

3 Far Far Rather High

4 Close Medium Medium High

5 Medium Medium Medium

6 Far Medium Medium Low

7 Close Close Rather Low

8 Medium Close Low

9 Far Close Very Low

this drawback, with the normalization of that sum by means

of division of local_distance by the number of nodes

that are within the specific radius of transmission.

CH_dispersion, is the sum of distances between the

candidate node and the cluster-heads within a radius of trans-

mission. This variable is also normalized. However, the greater

the sum, the higher the chance to coordinate the cluster.

Besides promoting a good distribution of cluster-heads, it

balances the transmission loads and the network processing.

The estimation of positions and distances is done by the

received signal strength intensity (RSSI), at a set-up phase.

Since ηs is the combination of two variables based on

imprecise measurements of distance, the fuzzy logic is well

suited to represent its final processing. Fuzzy logic is a

mathematical tool that relies on purely qualitative variables,

in contrast to the quantitative nature of crisp values [10]. In

WSNs, besides the uncertainty of distance measurements by

RSSI, the positioning of the sensor nodes can be subjected to

little changes, what can be compensated in the fuzzy inference

operation.

In fuzzy inference systems decisions are based on fuzzy

IF-THEN rules, linguistic variables and logical operators. The

fuzzy rules used in LF-Ant, to generate the value of the heuris-

tic information, are presented in Table I. Once the values of

local_distance and CH_dispersion becomes small

and great values, respectively, the fuzzy heuristic information

presents the higher values, and thus, the greater chances of

elect a cluster-head.

IV. COOPERATIVE MODULATION DIVERSITY

Concerning the biological behaviour of ants, a group trans-

mission can be seen as an efficient way to carry a large prey

to the nest, since ants working as a group can carry close to

ten times the carrying capacity of a solitary ant [11].

In LF-Ant, that collaboration between ants is translated

in to the use of a cooperative technique between the nodes

to mitigate the effects of the channel fading. Cooperative

modulation diversity (CDM) exploits diversity gain in a system

if each component of the transmitted signal is affected by inde-

pendent channel fading. Furthermore, to achieve the maximum

diversity gain, any two signal points in the system constellation

must have the maximum number of distinct components. The

collaborative operation is the major difference between CDM

and the classical modulation diversity. Moreover, the concept

of vice cluster-head is provided to support that operation in

multi-hop WSNs. CDM is invoked in an adaptive manner,

since it is needed only if transmission errors occur. The next

transmission hop from a source cluster-head can be performed

in two stages: broadcast and retransmission.

A. Broadcast Stage

In the broadcast stage, a source cluster-head forwards its

clustered message to another one, i.e., the closest neighbour

cluster-head and the next hop in the routing process towards

the sink node. The relay candidate, indicated by the relaying

threshold, receives the transmitted packet due to the broadcast

nature of the wireless channel. The remaining cluster sensor

nodes activate the sleep mode and save energy. The packet is

transmitted by a conventional QPSK modulation scheme. The

modulated signal is given by [6]

s(t) = A+∞∑

n=−∞

anp(t− nTs) cos(2πfct)

+A+∞∑

n=−∞

bnp(t− nTs) sin(2πfct), (6)

in which

an, bn = ±1 with equal probability

p(t) =

{

1, 0 ≤ t ≤ Ts

0, elsewhere(7)

for a carrier frequency, fc, and a carrier amplitude, A. The

transmitted packet has a CRC attached and the receiver (either

a neighbour cluster-head, or the relay candidate, or the sink

node) detects it. An acknowledgement is sent back to the

source cluster-head. If the packet is correctly detected by the

receiver, not necessarily the relay candidate, the source cluster-

head remains transmitting new packets and the previous pro-

cess is repeated. Otherwise, the retransmission stage begins.

B. Retransmission Stage

If the relay candidate receives the packet correctly, it

becomes a vice cluster-head and, jointly with the source

cluster-head, they retransmit the packet using the cooperative

modulation diversity. Otherwise, a simple QPSK transmission

is used again, the relay candidate activates the sleep mode and

saves energy. The retransmissions continue until the packet

is successfully delivered, or the number of retransmissions

exceeds Nmaxr , which is a preset parameter indicating the

maximum number of retransmissions allowed per packet. The

value of Nmaxr can depend on the application.

In CDM, if a QPSK constellation is rotated by a certain

angle, a kind of redundancy between the two quadrature

channels is introduced and the system can take advantage of

the derived diversity. Then, both the source and vice cluster-

heads rotate the constellation by an angle θ

s(t) = A+∞∑

n=−∞

xnp(t− nTs) cos(2πfct),

+A+∞∑

n=−∞

ynp(t− nTs) sin(2πfct), (8)

291

in which

xn = an cos θ − bn sin θ,

yn = bn sin θ + bn cos θ.

The constant phase θ is selected in such a way that the squared

Euclidean distance between QPSK signal constellations is

maximized for both components, inphase and quadrature [6].

Quadrature components are generated and each component

is independently interleaved. The signal interleavers are chosen

such that after deinterleaving, the two components will be

independent. For simplicity, consider the interleaving of only

two symbols. The first symbol transmitted by the source

cluster-head has the quadrature component of the second

symbol. On the other hand, the vice cluster-head transmits

a symbol with the quadrature component of the first symbol.

It can be perceived that the nodes involved in the cooperative

modulation transmission send just half of the total information

amount, individually. The two components are then upcon-

verted to the carrier frequency and added, using the following

Expression

ss(t) = A+∞∑

n=−∞

xnp(t− nTs) cos(2πfct) (9)

+A+∞∑

n=−∞

yn−kp(t− nTs) sin(2πfct), (10)

in which k is an integer representing the time delay in number

of symbols introduced by the interleaving between the I and

Q components.

C. The Channel Model and the Decoding System

Consider a communication channel with frequency nonse-

lective slowly fading with a multiplicative factor representing

the effect of fading and an additive term representing the

Gaussian noise. The received signal is

r(t) = α(t)s(t) + n(t), (11)

in which α(t) is modelled as zero-mean complex Gaussian

process. The received signal, r(t), is first downconverted to

baseband. The obtained signal (equivalent lowpass) in one

signalling interval is

rl(t) = αne−jφnsl(t) + z(t), nTs ≤ t ≤ (n+ 1)Ts, (12)

in which z(t) represents the complex white Gaussian noise, αn

is the fading amplitude (considered constant over one symbol

interval), φn is the phase shift due to the channel fading, and

sl(t) corresponds to the equivalent low pass of the transmitted

signal s(t) [6]. With the phase shift estimation of the received

signal at the sink node, and after the demodulation, the

received vector is given by

r̃n = αnsn + zn, (13)

in which sn is the vector representation of the transmitted

signal at time nTs, and the elements of the complex vector

zn are independent identically distributed Gaussian random

variables with zero mean and variance N0/2.

The decoded vector at the sink node, after the deinterleaving

process, is

rn = αnxn + Re{zn}+ j[αn+kyn + Im{zn}] (14)

which is then processed using symbol-by-symbol detection.

The optimum demodulator computes the squared Euclidean

distance between the received vector and each of the four

signal vectors of the QPSK scheme and then decides in favor

of the one closest to rn [6].

V. SIMULATION RESULTS

In the simulations, the sensor network is composed by 100

nodes. The nodes are deployed randomly on an area of 50×50meters. The sink node is located at the coordinates x = 25 and

y = 150 meters. It is assumed that each node has an initial

energy of 3 mJ. The radio dissipates εelec = 50 nJ/bit to run

the transmitter or receiver circuitry and εfs = 10 pJ/bit/m2,

or εmp = 0.0013 pJ/bit/m4 for the transmitting amplifier to

achieve an acceptable Eb

N0

. Consider d0 as a specific threshold

distance, given by

d0 =

√

εfs

εmp

. (15)

Thus, to transmit a κ-bit message at a distance d using the

radio model, the radio spends [9]

ETx(κ, d) =

{

κ · (εelec + εfs · d2), if d ≤ d0κ · (εelec + εmp · d4), if d > d0

(16)

and to receive this message, the radio spends

ERx(κ) = εelec · κ. (17)

The performance evaluation was done by comparing simu-

lation results between LF-Ant and another cooperative cross-

layer design, SCA with LEACH. Both systems use a truncated

ARQ scheme and the simulations were processed in Matlab

7.

SCA with LEACH uses the LEACH protocol at the network

layer and the cooperative space-time block codes at the phys-

ical layer, with the adaptive invoking of cooperative diversity,

at the link layer, only if errors occur. Even with similar bit

error rate behaviour [5] of the modulation diversity, the amount

of encoded data transmitted is twice the original message. In

current simulations, two cooperative nodes are used in the

SCA with LEACH operation

Figure 1 presents the performance comparison related to

network lifetime, in which the number of rounds for the

last dead node is evaluated, as a function of the channel

SNR and Nmaxr . It can be noted that for all the propagation

conditions, the LF-Ant protocol presents better results than

SCA with LEACH, since the network lifetime is extended

for more rounds. Furthermore, related to Nmaxr , the LF-Ant is

very constant, on the contrary of SCA with LEACH, which,

mainly for poor propagation conditions, decreases the amount

of rounds, because more energy is spent in more allowed

retransmissions.

The performance evaluation related to the packet loss rate

as a function the channel SNR is presented in Figure 2. For

292

low values of SNR, the LF-Ant performance is similar to

that observed for the SCA with LEACH, because the channel

quality is severely degraded. However, beyond 12 dB, the

PLR measured in LF-Ant decreases more than using SCA

with LEACH, which also contributes to enhance the network

lifetime.

12345678910369121518212427300

100

200

300

400

500

600

700

800

Nr

max

Channel SNR (dB)

Am

ou

nt

of

rou

nd

s f

or

the

la

st

de

ad

no

de

LF−Ant

SCA with LEACH

100

200

300

400

500

600

700

Fig. 1. The lifetime performance comparison between the proposed LF-Antand another cross-layer design. In all cases, LF-Ant presents better results.

3 6 9 12 15 18 21 24 27 300

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Channel SNR (dB)

Packet

Lo

ss R

ate

(P

LR

)

LF−Ant

SCA with LEACH

Fig. 2. The packet loss rate comparison as a function of the channel quality.Under poor propagation conditions, the performance evaluation is similar, butbeyond 12 dB, LF-Ant presents lower values of PLR.

The simulations showed that the greatest contribution to

the performance enhanced is due to the LF-Ant clustering

protocol. However, the joint operation with CMD reinforces

the superiority of the cross-layer proposed scheme.

VI. CONCLUSIONS AND FUTURE RESEARCH

The authors presented the design and evaluation perfor-

mance of a novel cooperative cross-layer protocol for multi-

hop wireless sensor networks. The LF-Ant protocol is bi-

ologically inspired in the organized behaviour of ants, in

which a classical ACO system is adapted to the considered

WSN constraints. The main goal is the optimal election of

cluster-heads and controlling the process of possible further

cooperative retransmissions.

The performance evaluation was done by comparing simu-

lation results between LF-Ant and another cooperative cross-

layer design, SCA with LEACH. The first performance metric

evaluated was the lifetime as function of the channel SNR and

of the maximum number of retransmissions allowed by the

truncated ARQ scheme used. The other evaluated parameter

was the packet loss rate as a function of the channel SNR.

In both evaluations, LF-Ant overcomes SCA with LEACH,

increasing the network lifetime and decreasing the PLR. That

superiority can be explained by the energy efficiency in

running the cooperative diversity by the LF-Ant, since the

nodes involved in the cooperation need to transmit just half of

the total symbols, individually. Furthermore, the bio-inspired

cluster-head election takes into account variables related to

residual energy, quality and consumption of previous transmis-

sions, and deals with the uncertainty of distance estimations,

by a fuzzy heuristic information modelling.

Future research includes the study of the interleaving depth

impact on the network lifetime and transmission delay in the

LF-Ant operation.

ACKNOWLEDGEMENT

The authors would like to thank Iecom, UFCG, CNPq for

supporting this research.

REFERENCES

[1] S. Selvakennedy, S. Sinnappan, and Yi Shang. A biologically-inspiredclustering protocol for wireless sensor networks. Computer Communi-

cations, 30:2786–2801, October 2007.[2] Y. Chen, J. Zhang, and I. Marsic. Link-layer-and-above diversity in mul-

tihop wireless networks. IEEE Communications Magazine, 47(2):118–124, February 2009.

[3] M. Dorigo and G. Caro. Ant colony optimization: a new meta-heuristic.In Proceedings of the Congress on Evolutionary Computation (CEC’99),1999.

[4] M. P. Sousa, R. F. Lopes, W. T. A. Lopes, and M. S. Alencar. Low-energy selective cooperative diversity with ARQ for wireless imagesensor networks. In IEEE 72nd Vehicular Technology Conference Fall

(VTC - Fall), pages 1–5, sept. 2010.[5] M. P. Sousa, A. Kumar, M. S. Alencar, and W. T. A. Lopes. Scalability

in an adaptive cooperative system for wireless sensor networks. InIEEE International Conference on Ultra Modern Telecommunications

Workshops (ICUMT 2009), St. Petersburg, Russia, 2009.[6] S. B. Slimane. An improved PSK scheme for fading channels. IEEE

Transactions on Vehicular Technology, 47(2):703 –710, May 1998.[7] S. A. Ahmadzadeh, S. A. Motahari, and A. K. Khandani. Signal

space cooperative communication. IEEE Transactions on Wireless

Communications,, 9(4):1266–1271, april 2010.[8] G. Caro and M. Dorigo. AntNet: distributed stigmergetic control for

communications networks. Journal of Artificial Intelligence Research,9:317–365, December 1998.

[9] J. Kim, S. Park, Y. Han, and T. Chung. CHEF: Cluster Head Elec-tion mechanism using Fuzzy logic in Wireless Sensor Networks. In10th International Conference on Advanced Communication Technology

(ICACT’08), pages 654–659, 2008.[10] Lotfi A. Zadeh. Fuzzy sets. Information and Control, 8(3):338–353,

1965.[11] G. Montemayor and J. Wen. Decentralized collaborative load transport

by multiple robots. In IEEE International Conference on Robotics and

Automation (ICRA’05), pages 372–377, April 2005.

293